Uniqueness of minimal projections in smooth matrix spaces (Q5929853)

Let \(M(n,m)\) be the linear space of matrices of \(n\) rows and \(m\) columns endowed with the norm \(\|\cdot\|,\) and let \(M(n,1),(M(1,m))\) be the space of matrices in \(M(n,m)\) with constant rows, (constant columns), respectively. A description formula for a minimal projection \(Q:X\rightarrow W :=M(n,1)+M(1,m)\) is known [see \textit{E. W. Cheney} and \textit{W. A. Light}, Approximation theory in tensor product spaces (1985; Zbl 0575.41001]. The main result: If the norm \(\|\cdot\|\) fulfills some conditions then the quoted \(Q\) is the unique minimal projection from \(X\) to \(W.\) For instance, if the norm is \(\|\cdot\|_p,\) for \(p\in (1,\infty)\) then \(Q\) is unique and for \(\|\cdot\|_1\) and \(\|\cdot\|_\infty\) the uniqueness is not assured. The author gives also sufficient conditions on the \(N\)-function \(\varphi\) such that the Orlicz or the Luxemburg norm \(\|\cdot\|_\varphi\) implies the unicity of \(Q.\)